高效快速的随机值脉冲噪声去除算法  

Algorithm of the effective and quick removal of random impulse noises

作　　者：陈庆强 蔡文培 邹复民[2] CHEN Qingqiang;CAI Wenpei;ZOU Fumin(School of Information Science and Engineering,Fujian University of Technology,Fuzhou 350118,China;Fujian Key Laboratory for Automotive Electronics and Electric Drive,Fujian University of Technology,Fuzhou 350118,China)

机构地区：[1]福建工程学院信息科学与工程学院,福建福州350118 [2]福建工程学院福建省汽车电子与电驱动技术重点实验室,福建福州350118

出　　处：《福建工程学院学报》2018年第3期242-247,共6页Journal of Fujian University of Technology

基　　金：国家自然科学基金资助项目(61304199)

摘　　要：提出一种既能快速去除图像随机脉冲噪声又能较好地保留边缘细节信息的一种新方法。该方法首先利用图像局部灰度相似性特征,对于任一像素,根据其与邻域内像素相近的个数和与其相近像素本身在邻域内的相似情况,将各像素分为噪声点、疑似噪声点和信号点,对疑似噪声点根据其是否为邻域内的极值将其分为噪声点和信号点。对于信号点不做任何处理,而对于噪声点则按照一种基于欧拉距离的自适应加权均值滤波算法进行处理。实验结果表明,算法能够快速高效地滤除随机脉冲噪声,且无需人为修改相关参数和"门坎"值,综合性能优良,特别适用于对实时性要求较高的图像处理系统。A new algorithm was put forward that can remove random impulse noises quickly and retain edge details well. The method first utilized the local gray-scale similarity of the image. According to the number of pixels in the neighbourhood of a certain pixel and its similarity with its neighbouring ones,the pixels could be classified into noise pixels,suspected noise pixels and signal pixels. The suspected noise pixels were then divided into poise pixels and signal pixels according to whether they were the extremum in the neighborhood. Signal pixels would not be processed,while noise pixels were processed with an adaptive weight-mean filtering algorithm based on Euler distance. Experimental results indicate that the proposed algorithm can filter out random impulse noises quickly and effectively. Moreover, it does not need manual adjustment of the parameters and thresholds. The proposed method achieves good comprehensive performances,and it is particularly suitable for image processing systems with high real-time requirements.

关 键 词：随机值脉冲噪声 噪声检测 欧拉距离 图像滤波 自适应 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]